Lecture26: Final Review · 2017-12-10 · Key idea: advertise the entire path Distance vector: send distance metricper destination Path vector: send the entire pathfor each destination

CSE 123: Computer NetworksAlex C. Snoeren

Lecture 26:Final Review

Project 2 due tonight!

Exam Overview● Focus on topics since the midterm

◆ But everything is fair game…

● Roughtly the same style & length as midterm◆ It won’t take the whole time◆ We’ll spend the first few minutes of the exam period awarding

the Espresso Prize

● As with the midterm, you can bring crib sheet◆ One double-sided 8x5.x11” paper that you turn in

2CSE 123 – Lecture 26: Final Review

TCP/IP Protocol Stack

HTTP

TCP

IP

Ethernetinterface

HTTP

TCP

IP

Ethernetinterface

IP

IP

Ethernetinterface

Ethernetinterface

SONETinterface

SONETinterface

host host

router router

3

Application Layer

Transport Layer

Network Layer

Link Layer

CSE 123 – Lecture 26: Final Review

Encapsulation via Headers● Typical Web packet

● Notice that layers add overhead◆ Space (headers), effective bandwidth◆ Time (processing headers, “peeling the onion”), latency

IP Hdr Payload (Web object)TCP Hdr HTTP HdrEthernet Hdr

Start of packet End of packetdatalink network transport application


Phy/(MAC)Link layer● Signal encoding

◆ Encode binary data from source node into signals that physical links carry

◆ Signal is decoded back into binary data at receiving node◆ Work performed by network adapter at sender and receiver

● Media access◆ Arbitrate which nodes can send frames at any point in time◆ Not always necessary; e.g. point-to-point duplex links


DNS ExampleHost at cis.poly.edu

wants IP address for gaia.cs.umass.edu

requesting hostcis.poly.edu

gaia.cs.umass.edu

root DNS server

local DNS serverdns.poly.edu

1

23

4

5

6

authoritative DNS serverdns.cs.umass.edu

78

TLD DNS server


10.0.0.4

10.0.0.2

10.0.0.3

S: 10.0.0.4:3345D: 132.239.8.45:80

110.0.0.1

138.76.29.7

1: host 10.0.0.4 sends packet to 132.239.8.45:80

NAT translation tableWAN side addr LAN side addr

138.76.29.7:5001 10.0.0.4:3345…… ……

S: 132.239.8.45:80 D: 10.0.0.4:3345 4

S: 138.76.29.7:5001D: 132.239.8.45:802

2: NAT routerchanges packetsource addr from10.0.0.1:3345 to138.76.29.7:5001,updates table

S: 132.239.8.45:80 D: 138.76.29.7:5001 3

3: Reply arrivesdest. address:138.76.29.7:5001

4: NAT routerchanges packetdest addr from138.76.29.7:5001 to 10.0.0.4:3345

NA(p)T Example


● How to choose best path?◆ Defining “best” can be slippery

● How to scale to millions of users?◆ Minimize control messages and routing table size

● How to adapt to failures or changes?◆ Node and link failures, plus message loss

8

Routing Challenges


● Source routing◆ Complete path listed in packet

● Virtual circuits◆ Set up path out-of-band and store path identifier in routers◆ Local path identifier in packet

● Destination-based forwarding◆ Router looks up address in forwarding table◆ Forwarding table contains (address, next-hop) tuples

9

Forwarding Options


● Two phases◆ Reliable flooding

» Tell all routers what you know about your local topology◆ Path calculation (Dijkstra’s algorithm)

» Each router computes best path over complete network

● Motivation◆ Global information allows optimal route computation◆ Straightforward to implement and verify

10

Link-state Routing


Dijkstra’s Shortest Path Tree● So you have all of these LSPs. Now what?● Graph algorithm for single-source shortest path tree

(find best route to all nodes)

S ß {}Q ß <remaining nodes keyed by distance>While Q != {}

u ß extract-min(Q) u = node with lowest costS ß S plus {u}Within Q:

for each node v adjacent to u“relax” the cost of v is it cheaper to go

through u?

ß u is done

11

Iterative, asynchronous: each local iteration caused by:

● Local link cost change ● Distance vector update message

from neighbor

Distributed:● Each node notifies neighbors

when its DV changes● Neighbors then notify their

neighbors if necessary

wait for (change in local link cost or message from neighbor)

recompute estimates

if distance to any destination has changed, notify neighbors

Each node:

Distance Vector Algorithm


● Define distances at each node X◆ dx(y) = cost of least-cost path from X to Y

● Update distances based on neighbors◆ dx(y) = min {c(x,v) + dv(y)} over all neighbors V

32

2

1

14

1

4

5

3

u

v

w

x

y

z

s

t du(z) = min{c(u,v) + dv(z), c(u,w) + dw(z)}

Bellman-Ford Algorithm


1A CB

23 2

1A CB3 4

Update 3

1A CB

Update 4

5 4

Etc…

Distance to C

Problem: Counting to Infinity


● Hold downs◆ As metric increases, delay propagating information◆ Limitation: Delays convergence

● Loop avoidance◆ Full path information in route advertisement◆ Explicit queries for loops

● Split horizon◆ Never advertise a destination through its next hop

» A doesn’t advertise C to B◆ Poison reverse: Send negative information when advertising a

destination through its next hop» A advertises C to B with a metric of ¥» Limitation: Only works for “loop”s of size 2

Mitigation Strategies


● Internet is divided into Autonomous Systems◆ Distinct regions of administrative control◆ Routers/links managed by a single “institution”◆ Service provider, company, university, …

● Hierarchy of Autonomous Systems◆ Large, “tier-1” provider with a nationwide backbone◆ Medium-sized regional provider with smaller backbone◆ Small network run by a single company or university

● Interaction between Autonomous Systems◆ Internal topology is not shared between ASes◆ … but, neighboring ASes interact to coordinate routing

Autonomous Systems


● Border routers summarize and advertise their routes to external neighbors and vice-versa

◆ Border routers apply policy

● Internal routers can use notion of default routes

● Core is default-free; routers must have a route to all networks in the world

● But what routing protocol?

R1

Autonomous system 1R2

R3

Autonomous system 2R4

R5 R6

AS1

AS2

Border router

Border router

Inter-domain Routing


● Extension of distance-vector routing◆ Support flexible routing policies◆ Avoid count-to-infinity problem

● Key idea: advertise the entire path◆ Distance vector: send distance metric per destination◆ Path vector: send the entire path for each destination

3 2 1

d

“d: path (2,1)” “d: path (1)”

data traffic data traffic

Path-vector Routing


● Destination prefix (e.g., 128.112.0.0/16)● Route attributes, including

◆ AS path (e.g., “7018 88”)◆ Next-hop IP address (e.g., 12.127.0.121)

19

AS 88Princeton

128.112.0.0/16AS path = 88Next Hop = 192.0.2.1

AS 7018AT&T

AS 11Harvard

192.0.2.1

128.112.0.0/16AS path = 7018 88Next Hop = 12.127.0.121

12.127.0.121

A Simple BGP Route

● Neighboring ASes have business contracts◆ How much traffic to carry◆ Which destinations to reach◆ How much money to pay

● Common business relationships◆ Customer-provider

» E.g., Princeton is a customer of USLEC» E.g., MIT is a customer of Level3

◆ Peer-peer» E.g., UUNET is a peer of Sprint» E.g., Harvard is a peer of Harvard Business School

Business Relationships


Functional architecture Reservation/Admission

Control

Routing Protocols

Routing Table

ClassificationRules

Firewall

PacketClassification Switching

ForwardingTable

OutputScheduling

Control Plane• Complex• Per-control action• May be slow

Data plane• Simple• Per-packet• Must be fast

21

Interconnect architecture● Input & output connected via switch fabric

● Kinds of switch fabric◆ Shared Memory◆ Bus◆ Crossbar

● How to deal with transientcontention?◆ Input queuing◆ Output queuing

Input OutputSwitch


IQ + Virtual Output Queuing● Input interfaces buffer packets

in per-output virtual queues

● Pro◆ Solves blocking problem

● Con◆ More resources per port◆ Complex arbiter at switch◆ Still limited by input/output

contention (scheduler)

Input OutputSwitch


● Buffer management: which packet to drop when? ◆ We only have finite-length queues

● Scheduling: which packet to transmit next?

1

2

Scheduler

flow 1

flow 2

flow n

Classifier

Buffer management

Key Router Challenges


Min threshMax thresh

Average Queue Length

minth maxth

maxP

1.0

Avg queue length

P(drop)

RED Operation


● Parameters◆ r – average rate, i.e., rate at which tokens fill the bucket◆ b – bucket depth (limits size of burst)◆ R – maximum link capacity or peak rate (optional parameter)

● A bit can be transmitted only when a token is available

r bps

b bits

≤ R bps

regulatortime

bits

b·R/(R-r)

slope R

slope r

Maximum # of bits sent

b/(R-r)

Token Bucket Basics


● Rough definition: “When an increase in network load produces a decrease in useful work”

● Why does it happen?◆ Sender sends faster than bottleneck link speed◆ Packets queue until dropped◆ In response to packets being dropped, sender retransmits◆ All hosts repeat in steady state…

27

Congestion Collapse


Network Load

Thro

ughp

utLa

tenc

y

“Knee”“Cliff”

Congestion collapse

● Congestion avoidance: try to stay to the left of the knee● Congestion control: try to stay to the left of the cliff

28

Proactive vs. Reactive

Loss due toCongestion


● Window-based congestion control◆ Unified congestion control and flow control mechanism◆ rwin: advertised flow control window from receiver◆ cwnd: congestion control window

» Estimate of how much outstanding data network can deliver in a round-trip time

◆ Sender can only send MIN(rwin,cwnd) at any time

● Idea: decrease cwnd when congestion is encountered; increase cwnd otherwise◆ Question: how much to adjust?

29

Basic TCP Algorithm


cwnd=1

cwnd=2

cwnd=4

cwnd=8

Sender Receiver

0

50

100

150

200

250

300

0 1 2 3 4 5 6 7 8round-trip times

cwnd

30

TCP Slow Start


Slow Start + Congestion Avoidance

0

2

4

6

8

10

12

14

16

18

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40

round-trip times

cwnd

Timeout

ssthresh

Slow start

Congestionavoidance

31

Basic TCP Mechanisms


● Maintain a queue for each flow◆ What is a flow?

● Implements max-min fairness: each flow receives min(ri, f) , where

◆ ri – flow arrival rate◆ f – link fair rate (see next slide)

● Weighted Fair Queuing (WFQ) – associate a weight with each flow to divvy bandwidth up non-equally

Fair Queuing


0 2 104 6 8

0 2 104 6 8

● Select the first packet that finishes in the fluid flow system

Servicein fluid flow

system

Packetsystem

Packet-Based WFQ


Signals and Channels● A signal is some form of energy (light, voltage, etc)

◆ Varies with time (on/off, high/low, etc.)◆ Can be continuous or discrete

● A channel is a physical medium that conveys energy◆ Any real channel will distort the input signal as it does so◆ How it distorts the signal depends on the signal


Carrier Signals● Baseband modulation: send the “bare” signal

◆ E.g. +5 Volts for 1, -5 Volts for 0◆ All signals fall in the same frequency range

● Broadband modulation◆ Use the signal to modulate a high frequency signal (carrier).◆ Can be viewed as the product of the two signals

35

Ampl

itude

Signal CarrierFrequency

Ampl

itude

ModulatedCarrier


Forms of Digital Modulation

36

Input Signal

Amplitude Shift Keying(ASK)

Frequency Shift Keying(FSK)

Phase Shift Keying(PSK)


Shannon’s Law● Shannon considered noisy channels and derived

C = B log (1 + S/N)

● Gives us an upper bound on any channel’s performance regardless of signaling scheme

● Old school modems approached this limit◆ B = 3000Hz, S/N = 30dB = 1000◆ C = 3000 x log(1001) =~ 30kbps◆ 28.8Kbps – anyone remember dialup?


Nyquist Sampling Rate● Sampling at the correct rate (2f) yields actual signal

◆ Always assume lowest-frequency wave that fits samples

● Sampling too slowly yields aliases


Clock Recovery● Using a training sequence to get receiver lined up

◆ Send a few, known initial training bits◆ Adds inefficiency: only m data bits out of n transmitted

● Need to combat clock drift as signal proceeds◆ Use transitions to keep clocks synched up

● Question is, how often do we do this?◆ Quick and dirty every time: asynchronous coding◆ Spend a lot of effort to get it right, but amortize over lots of

data: synchronous coding


Manchester Encoding● Signal to Data

◆ XOR NRZ data with senders clock signal◆ High to low transition ð 1◆ Low to high transition ð 0

● Comments◆ Solves clock recovery problem◆ Only 50% efficient ( ½ bit per transition)◆ Still need preamble (typically 0101010101… trailing 11 in Ethernet)

Bits 0 0 1 0 1 1 1 1 0 1 0 0 0 0 1 0

NRZ

Clock

Manchester


Fixed PartitioningFDMA

TDMA

CDMA

pow

er

pow

er

pow

er

Courtesy Takashi Inoue


Aloha● Designed in 1970 to support wireless data connectivity

◆ Between Hawaiian Islands—rough!

● Goal: distributed access control (no central arbitrator)◆ Over a shared broadcast channel

● Aloha protocol in a nutshell:◆ When you have data send it◆ If data doesn’t get through (receiver sends acknowledgement)

then retransmit after a random delay◆ Why not a fixed delay?


Q: What is max fraction slots successful?A: Suppose n stations have packets to send

◆ Each transmits in slot with probability p◆ Prob[successful transmission], S, is:

S = p (1-p)(n-1)

◆ any of n nodes:

S = Prob[one transmits] = np(1-p)(n-1)

(optimal p as n->infinity = 1/n)

= 1/e = .37

At best: channelused for useful transmissions 37%of time!

offered load = n X p0.5 1.0 1.5 2.0

0.1

0.2

0.3

0.4

Pure Aloha

Slotted Aloha

Slotted Aloha Efficiency


● How can A know that a collision has taken place?◆ Worst case:

» Latency between nodes A& B is d» A sends a message at time t and B sends a message at t + d – epsilon (just before

receiving A’s message)◆ B knows there is a collision, but not A… A must keep transmitting until it

can tell if a collision occurred◆ How long? 2 * d

● IEEE 802.3 Ethernet specifies max value of 2d to be 51.2us◆ This relates to maximum distance of 2500m between hosts◆ At 10Mbps it takes 0.1us to transmit one bit so 512 bits take 51.2us to send◆ So, Ethernet frames must be at least 64B (512 bits) long

» Padding is used if data is too small● Send jamming signal to insure all hosts see collision

◆ 48 bit signal

44

CSMA/CD


45

CSMA/CA● Cannot detect collision w/half-duplex radios

● Wireless MAC protocols often use collision avoidance techniques, in conjunction with a (physical or virtual) carrier sense mechanism

● Collision avoidance◆ Nodes negotiate to reserve the channel.◆ Once channel becomes idle, the node waits for a

randomly chosen duration before attempting to transmit.


46

A B C

● When A wants to send a packet to B, A first sends a Request-to-Send (RTS) to B

● On receiving RTS, B responds by sending Clear-to-Send (CTS), provided that A is able to receive the packet

● When C overhears a CTS, it keeps quiet for the duration of the transfer◆ Transfer duration is included in both RTS and CTS

RTS/CTS (MACA)


Parting thoughts…

● Good luck finishing up Project 2◆ Don’t use the late days…

● Please complete your CAPE survey online

● Final exam: TUESDAY 3-6 PM

● Good luck and have a great holiday break!


Documents

Lecture26: Final Review · 2017-12-10 · Key idea: advertise the entire path Distance vector: send distance metricper destination Path vector: send the entire pathfor each destination